Use this URL to cite or link to this record in EThOS:
Title: A phenomenological-connectionist theory of computational agency
Author: Joyce, Daniel W.
ISNI:       0000 0001 3593 3300
Awarding Body: University of Southampton
Current Institution: University of Southampton
Date of Award: 2001
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
This thesis presents a theory of computational agency. Computational agent theory differs from 'classical' artificial intelligence by committing to the view that a computational artifact is situated, and that its rationality is limited by the constraints of this 'situatedness'. Contempo- rary literature is surveyed and models of situated computational agency placed in their philo- sophical contexts. From this, a critical reconstruction of the notion of an agent is given from a phenomenological perspective. It is proposed that everyday routines of activity underpins agency and computational implementations of this substrate can take the form of connectionist networks. The proposal is tested in technical practice on two domains; agents for multimedia content- based navigation/retrieval, and a simulated environment which explores the key properties of the proposed phenomenological agent theory. Recent proposals for goal-directed behaviour in connectionist systems (largely from the cognitive and behavioural neurosciences) are critically evaluated, and integrated into an agent architecture. This results in an architecture utilising suit- ably controlled reinforcement learning. The architecture implemented is then evaluated against the agent theory, and examples of 'routine behaviour' analysed in stationary and non-stationary environments. Semiotic analyses are then proposed as an alternative theory of representation, as they are compatible with, and simultaneously possess explanatory power at a level beneath, the usual sentential/propositional level. The thesis contributes a phenomenological theory of agency and gives examples of its influence on technical practice. Outlines of connectionist architectures are presented, imple- mented, and evaluated with respect to the agent theory proposed.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available
Keywords: Neural networks; Architectures